Material Extrusion Additive Manufacturing of Poly(Lactic Acid): Influence of infill orientation angle

(1)

Additive Manufacturing 59 (2022) 103079

Available online 12 August 2022

Contents lists available atScienceDirect

Additive Manufacturing

journal homepage:www.elsevier.com/locate/addma

Research paper

Material Extrusion Additive Manufacturing of Poly(Lactic Acid): Influence of infill orientation angle

Wilco M.H. Verbeeten

^∗

, Miriam Lorenzo-Bañuelos

Structural Integrity Research Group, Universidad de Burgos, Avenida Cantabria s/n, E-09006, Burgos, Spain

A R T I C L E I N F O

Keywords:

PLA

Angle-ply laminate configuration

Anisotropic strain-rate dependent yield stress Ree–Eyring rate equation

A B S T R A C T

The effect that the infill orientation angle has on the strain-rate dependence of the yield stress for material extrusion additive manufactured (ME-AM) PolyLactic Acid (PLA) material was investigated. Symmetric angle- ply stacking sequences were used to produce ME-AM tensile test samples. Measured yield stresses were compensated for the voided structure, typical of ME-AM components. Furthermore, molecular orientation and stretch was macroscopically assessed by a thermal shrinkage procedure. Additionally, hot-press compression molded (CM ) samples were manufactured and mechanically characterized in uniaxial tensile and compression in order to determine the material’s isotropic bulk properties. Initial model parameters for the Ree–Eyring modification of the Eyring flow rule were determined using CM data. According to SEM fractography, all samples showed microscopically brittle fracture behavior. Notwithstanding, contrary to CM samples, ME-AM specimens showed macroscopically ductile stress–strain behavior and a transition from a regime with only a primary 𝛼-deformation process, at low strain rates, to a regime with 2 deformation processes (𝛼 + 𝛽), at high strain rates. These effects are an influence of the processing step and are attributed to the molecular orientation and stretch of the polymer chains, provoking anisotropic mechanical properties. As a consequence, a deformation-induced change of the Eyring rate constants is needed to adequately describe the strain-rate dependence of the ME-AM yield stress behavior, leaving the initial activation volumes unchanged. Taking this deformation-dependence of the rate constants into account, yield stresses as a function of infill orientation angle can be appropriately predicted.

1. Introduction

Material Extrusion Additive Manufacturing (ME-AM) [1], also known as Fused Deposition Modeling (FDM^®), Fused Filament Fab- rication (FFF), Fused Layer Modeling (FLM), or 3D printing, is one of the most widely used Additive Manufacturing (AM) techniques for polymer materials. It is attracting a lot of attention from both industry as well as research groups, as it can be effectively used for end-user manufacturing. Its popularity is due to a combination of low investment costs, availability of a wide range of materials, and ease for manufacturing [2,3]. Furthermore, components with good and near-bulk mechanical properties can be obtained, if proper processing parameters are chosen [4–7].

The equipment for ME-AM technology consists of a pinch roller mechanism that pushes a thermoplastic polymer filament through a heated liquefier. The molten polymer is consequently extruded through a nozzle, and deposited in a controlled manner onto a (heated) build platform or an already deposited and solidified layer. In such a way, complex 3D products can be constructed layer by layer. During the

∗ Corresponding author.

E-mail address: [email protected](W.M.H. Verbeeten).

ME-AM process, material elements experience time–temperature (initial fast cooling combined with successive heating cycles) and time–

strain profiles [8–14], which are significantly different from more conventional methods, e.g. injection or compression molding tech- niques. Both profiles are influenced by the chosen printing process parameters [15,16], and provoke specific macroscopic deformation behavior as measured in macroscopic characterization tests (e.g. tensile, compression, shear, torsion, bending, creep, fatigue, impact). The time–

temperature profile affects bonding between the extruded filament and the already cooled adjacent strands though the wetting and molecular diffusion process driven by reptation [11,12,17]. Shear effects in the nozzle and the curvature between nozzle and deposited filament can lead to orientation and stretch of the polymer chain [11,13].

For amorphous polymers, frozen-in flow-induced molecular orientation can emerge [16,18]. While for semi-crystalline polymers, it can cause oriented crystalline microstructures (i.e. shish-kebab morphology) due to flow-induced crystallization [19]. For both type of polymers, this produces local anisotropy. Hence, the resulting end-product is a locally

https://doi.org/10.1016/j.addma.2022.103079

Received 4 May 2022; Received in revised form 2 August 2022; Accepted 7 August 2022

(2)

Unidirectional printed Acrylonitrile–Butadiene–Styrene (ABS) specimens were used by Rodríguez et al. [21], i.e. samples where the infill strand deposition angle of all layers is in the same direction. They showed that the yield strength as a function of strand orientation could be satisfactorily predicted by the Azzi and Tsai failure theory [26]

for composite laminates. Lanzotti et al. [22] also used unidirectional samples. However, the layer thickness and number of perimeters used in their study on PolyLactic Acid (PLA) material were not constant for the different infill angles. Song et al. [5] and Afrose et al. [27]

also opted for unidirectional samples, both using a PLA material. Song et al. conducted tension and compression tests at two different strain rates and 3 orientations, as well as fracture experiments and compared the results with injection molded specimen [5]. While Afrose et al.

investigated the fatigue behavior at 3 orientations [27].

Since ME-AM has similarities with laminar composite structures, often theory, nomenclature, and notations for lamina stacking sequence from that research area are adopted. One of the most commonly used infill orientation angle configurations for 3D printing is a so-called orthogonal angle-ply laminate configuration [28]. Such a configuration has a repetition of one layer at a certain infill angle and the next at an angle perpendicular to that. A standard default configuration is indicated by the notation [45^◦/−45^◦], and consists of one layer at a 45^◦ angle and the next at a −45^◦angle, leading to a criss-cross raster laminate structure. Another often used orthogonal angle-ply configuration is [0^◦/90^◦]. Among others, Tymrak et al. [29] used these two configurations to investigate ABS and PLA materials, Ning et al. [30] for chopped carbon fiber-reinforced ABS, and Papon & Haque [31] for short carbon fiber-reinforced PLA. Several other research groups compared orthogo- nal angle-ply specimen with unidirectional samples, e.g., to investigate ABS material [32], fatigue properties of PLA material [6], or mechani- cal strength and fatigue properties for Ultem (i.e. PolyEtherImide (PEI)) and PolyEtherKetoneKetone (PEKK) materials [33].

A more complete set of orthogonal angle-ply specimen, i.e. [0^◦/90^◦], [15^◦/−75^◦], [30^◦/−60^◦], and [45^◦/−45^◦], was adopted by Li et al. [20]

and Dawoud et al. [23] for ABS material, and by Ahmed & Susmel [24]

to investigate PLA material. Messimer et al. [34] executed a very complete study for eleven polymer materials, using 7 orthogonal angle- ply sample sets ranging from 0^◦ to 90^◦ in 15^◦ increments for each material. Unfortunately, only dimensional error was measured, but no mechanical strength properties.

Another sort of orientation configuration often used in laminar composite structures, is the (anti)symmetric angle-ply configuration. These specimen consist of an odd or even number of identical orthotropic layers oriented alternately at angles 𝜃 and −𝜃 [28,35], indicated as [𝜃/−𝜃]. Kain et al. [25] used 7 different angle-ply samples from 0^◦to 90^◦at 15^◦ intervals for their research on PLA/wood specimen. Note that this is a different set of samples than the orthogonal angle-ply configuration applied by e.g. Messimer et al. [34].

Occasionally, pseudo-isotropic samples are used for investigation purposes [36], e.g. by using a [0^◦/90^◦/45^◦/−45^◦] infill orientation stacking sequence. In laminar composite terms, such a configuration is also indicated as quasi-isotropic laminates [28]. Mechanical properties in the XY-plane are practically independent of the load direction for

behavior. This not only holds for short-term experiments (e.g. uniaxial tensile or compression tests), but also for long-term failure behavior, as manifested in creep and fatigue measurements [41–43].

The present study is focused on the effect that the infill orientation angle has on the strain-rate dependence of the yield stress for a PolyLactic Acid (PLA) material. Only samples oriented in the printer’s XY-plane are considered. Although the objective is somewhat similar to a previous study [7], the present research includes several novelties:

(i) Hot-press compression molded (CM ) samples are manufactured and mechanically characterized in both uniaxial tensile and compression tests, in order to determine the PLA material’s isotropic bulk refer- ence properties. By measuring in more than one loading geometry (i.e. uniaxial tensile and compression), a more complete and precise set of parameters for an Eyring-type flow rule [37,44] can be established [45]. For instance, thermorheologically complex behavior can be more easily detected in uniaxial compression tests [46]. Besides, an initial reference set for the Eyring-type flow rule sheds light on the polymer’s viscoelastic behavior [39], and how the ME-AM process influences that viscoelastic behavior.

(ii) Samples fabricated by the ME-AM technique are measured at more strain rates in a broad strain-rate spectrum ( ̇𝜀 = 10⁻⁵ − 10⁻¹s⁻¹). The advantage of measuring more strain rates is that transitions due to the appearance of an additional molecular relaxation process become more apparent [45,47].

(iii) An ample range of 7 different symmetric angle-ply stacking sequences (from 0^◦ to 90^◦ at 15^◦intervals) is used to produce ME-AM sample sets, similar to what was used by Kain et al. [25] in their study for wood-filled PLA. This is significantly more compared to the two unidirectional sets (longitudinal [0^◦] and transverse [90^◦]) that were used in the previous study [7]. Hence, infill orientation effects become more visible.

(iv) Both CM and ME-AM sample sets are compared to each other and, additionally, to an ME-AM quasi-isotropic sample set with a [0^◦/90^◦/45^◦/−45^◦] stacking sequence. For all these sets, the Eyring- type flow rule is evaluated to predict yield stress performance.

As far as the authors know, the combination of these four aspects has not been published before by other research groups.

The results of the present study can aid to better understand the influence of the infill orientation angle of deposited polymer strands on mechanical properties. Furthermore, it can help to further develop predictive numerical tools that can quantitatively predict the performance of material extrusion additively manufactured components. Es- pecially since Eyring-type flow rules can be satisfactorily incorporated in accurate constitutive models for polymer materials [48–51].

2. Materials and methods 2.1. Material

The material used for the present study is a commercially available natural transparent Polylactide (PLA) filament (Orbi-Tech, Leichlingen, Germany) with a nominal diameter of 1.75 mm and a specific gravity of 1.25 g cm⁻³. Recommended nozzle and bed temperatures range from 195^◦C to 240^◦C and from 40^◦C to 60^◦C, respectively. All samples were fabricated from a single spool, and the filament was used as-received.

(3)

Fig. 1. (a)Tensile test specimen dimensions in mm. (b) Schematic of the angle-ply infill orientation stacking sequence. Gray arrows indicate force direction in uniaxial tensile tests.

Table 1

Processing parameters used to manufacture ME-AM tensile samples.

Processing parameter Value

Nozzle diameter [mm] 0.40

Extrusion width [mm] 0.30

Extrusion multiplier 1.00

Layer height [mm] 0.20

Number of perimeters 2

Fill percentage 100%

Outline overlap 80%

Extrusion temperature [^◦C] 205

Bed temperature [^◦C] 50

Printing speed 𝑣𝑝[mm s⁻¹] 35

Table 2

Lamina stacking sequence to manufacture different sets of ME-AM tensile samples.

Set nomenclature Infill orientation stacking sequence

PLAo00 [0^◦]

PLAo15 [15^◦/−15^◦]

PLAo30 [30^◦/−30^◦]

PLAo45 [45^◦/−45^◦]

PLAo60 [60^◦/−60^◦]

PLAo75 [75^◦/−75^◦]

PLAo90 [90^◦]

PLAoISO [0^◦/90^◦/45^◦/−45^◦]

2.2. Material processing

Tensile samples were manufactured on an open-source Prusa P3steel 3D printer (Orballo Printing C.B., Vigo, Spain) using a 0.4 mm nozzle size. Sample dimensions are given in Fig. 1(a). This tensile test specimen is based on Specimen type 1BA according to the ISO 527-2 norm, but adapted to avoid fracture in the fillet [7]. The dimensions were chosen in such a way that they could also be machined from compression molded square plates.

An STL file of the tensile sample was imported in the Simplify3D slicing software (version 3.1.1). The printing parameters used to produce all ME-AM sample sets are given inTable 1. Note that an outline overlap of 80% was used to produce the test specimens. This affects only one of the perimeter outlines, and was chosen in order to avoid large gaps in the curved parts of the samples. It could, however, introduce additional residual stresses in the samples’ edges.

Sets of ME-AM samples with different infill orientation angles were produced, using a symmetric angle-ply stacking sequence. In total, 7

sets of angle-ply samples from 0^◦to 90^◦at 15^◦intervals were printed.

Additionally, a set of quasi-isotropic samples with a [0^◦/90^◦/45^◦/−45^◦] stacking sequence was manufactured. InTable 2, the stacking sequence of the different sets are indicated and a schematic is given inFig. 1(b).

The reference line to indicate the infill angle is the main axis of the tensile specimen, and coincides with the sample loading direction. An image of the printed and tensile tested (at ̇𝜀 = 10⁻³ s⁻¹) samples for different configurations is shown inFig. 2. After given the right processing parameters and stacking sequence to the test specimen, the sample was copied 18 times and distributed evenly on the printer’s XY plane. Thus, a total of 8 sets were printed, each set consisted of 18 samples and was manufactured with a single G-code file.

2.3. Hot-press compression molding

In order to obtain test specimen with isotropic bulk properties, i.e.specimen with a minimum of molecular orientation, samples were extracted from hot-press compression molded (CM ) plates. During the initial phase of compression molding, the material was heated and able to relax its possible residual orientation, leading to isotropic samples at the end of the fabrication cycle. This contrary to injection molding, where molecular orientation can be introduced during processing [52–

54]. The hot-press compression molding technique was used to produce both isotropic tensile and compression test samples.

For the tensile test specimen, square plates with dimensions of 100×

100 × 3mm³ were compression molded in a hot press. Approximately 45gr of chopped filament strands were preheated in the mold at 210^◦C for 15 min. Subsequently, the material was compressed during 5 min in successive steps of 5 bar by increasing the pressure from 5 bar to 20 bar and allowing degassing by releasing pressure between steps.

Next, cooling to room temperature was established by placing the mold into a cold press at a moderate pressure of 10 bar during another 5 min. From these square plates, tensile test samples were cut with a LASER PC60/40 laser cutting machine (PerezCamps, Viladecans, Spain) at 48 W power and 1000 mm min⁻¹speed with the dimensions as given inFig. 1(a). Sample edges were straight and smooth and showed no excessive melting or any burn marks after visual inspection. However, thermal effects due to laser cutting cannot completely be excluded.

For the compression test specimen, a square plate with dimensions of 100×100×8 mm³was produced using the same compression molding procedure. Here, approximately 120 gr of chopped filament strands were used. From the square plate, cylindrical compression test samples

(4)

Fig. 2. Image of tensile tested ME-AM samples with different stacking sequences.

were carefully turned on a CNC lathe in order to maintain a high level of parallelism and perpendicularity. The test specimens had a height and diameter of 6 mm.

2.4. Mechanical characterization

Uniaxial tensile tests were performed on a MTS Criterion C43.104 universal test system with a 10 kN load cell. All experiments were measured at room temperature (23^◦C). Constant nominal strain rates were applied in the range from 10⁻⁵s⁻¹ to 10⁻¹ s⁻¹. Three samples were used for every single strain rate, obtaining good repeatability and low standard deviations (see alsoTable A.2). As is common for polymers, the first maximum stress value in the engineering stress–

strain curves is considered as the polymer’s yield stress. Conversion from engineering to true yield stresses was accomplished by assuming that the material volume remains constant during uniaxial tensile testing [47,55]. This is standard in polymer engineering, and according to Govaert et al. [56] it introduces a small error of approximately 2%

compared to a compressible approach.

For uniaxial compression tests, the cylindrical samples were compressed between two parallel flat steel platens, mounted in the same MTS Criterion C43.104 universal test system. To reduce friction in order to avoid bulging or buckling of the samples, a thin PTFE skived film tape (3M 5480) was attached to the sample ends and the surface between steel platens and tape was lubricated with a PTFE spray.

Homogeneous deformation during the complete compression test range was observed, indicating that friction was sufficiently reduced. Con- stant nominal strain rates ranging from 3 ⋅ 10⁻⁴s⁻¹to 10⁻¹s⁻¹were applied at room temperature. Three test samples were used for every single strain rate. Since homogeneous deformation was obtained during testing, even at large strains, true stress–strain curves can be directly determined from the uniaxial compression tests under the assumption of incompressibility and applying a correction for the machine set-up stiffness.

2.5. Apparent density

Due to the fact that ME-AM parts generally have interstitial voids, its mechanical properties are affected by this voided structure. To compensate for this and provide a more fair comparison of the material behavior, the apparent densities of the resulting ME-AM samples were determined before performing mechanical characterization. Both mass and external volume were measured for every single sample to calculate the apparent density:

𝜌_app= 𝑚_sample

𝑉_sample . (1)

Furthermore, an approximation of the porosity of the samples (in percentage) was determined by using the material density as given by the filament provider, i.e. 𝜌_PLA= 1.25g cm⁻³:

Porosity =𝜌_PLA− 𝜌_app

𝜌_PLA ⋅ 100%. (2)

Hence, by multiplying measured yield stresses 𝜎_𝑦with the material’s reference density and divide it by the apparent density, void corrected yield stresses 𝜎_{𝑦,𝑣𝑐}can be calculated:

𝜎_{𝑦,𝑣𝑐}= 𝜎_𝑦⋅𝜌_PLA

𝜌_app . (3)

This calculation is done for every single sample separately.

2.6. Thermal shrinkage

For assessment of molecular orientation and stretch, a qualitative macroscopic measurement procedure was applied that is based on thermal shrinkage [16]. It consisted of measuring the ME-AM sample dimensions at room temperature. Next, the ME-AM samples were heated in an oven at 100^◦C for six hours. As determined using DSC measurements on the same material by Verbeeten et al. [7], this is a temperature well above the glass transition temperature (𝑇_𝑔 = 66^◦C) and below the cold crystallization temperature (𝑇_𝑐 = 122^◦C). At this temperature and during this time, the molecular chains are able to relax. Then, the samples were cooled back to room temperature and their dimensions were measured again. Expansion(+)/contraction(−) can be determined from dimensional differences. Three samples for each printed set were used for this thermal shrinkage procedure. In order to determine if and to what extent cold crystallization affected the dimensional shrinkage, Verbeeten et al. [7] performed DSC measurements on samples, which were elaborated at various printing velocities and two different raster angles. Both recently printed samples and samples that had undergone thermal shrinkage were measured. Crystallinities for the untreated ME-AM samples were below 2%. All thermally treated samples had similar crystallinities around 38%, but significantly different dimensional changes. Furthermore, all samples demonstrated dimensional contraction in the strand directions and expansion in the direction perpendicular to the strand deposition. Hence, it was concluded from the combination of thermal shrinkage and DSC measurements, that cold crystallization had a similar shrinkage effect on all samples, and that the variations in dimensional changes were indicative to the degree of molecular chain orientation and stretch due to the processing phase [7].

2.7. Scanning Electron Microscope fractography

Following uniaxial tensile tests, the fracture surface of several samples were observed using a JEOL JSM-6460LV scanning electron microscope (SEM). The samples were prepared by sputter-coating with an Au-coating. Samples were observed under vacuum and an accelerating voltage of 20.00 kV.

(5)

Table 3

Definitions of the equivalent (plastic) shear rate, ̇̄𝛾, shear stress, ̄𝜏, and hydrostatic pressure, 𝑝, expressed in components of the deformation and stress tensor. Explicit expressions are given for tension, compression, and shear.

Definition Tension Compression Shear

̇̄𝛾 =√ 2(

̇𝜀²

11+ ̇𝜀²₂₂+ ̇𝜀²₃₃+ 2 ̇𝜀²₁₂+ 2 ̇𝜀²₁₃+ 2 ̇𝜀²₂₃) √

3 ̇𝜀 √

3| ̇𝜀| ̇𝛾

̄ 𝜏=

√

1 6

[(𝜎₁₁− 𝜎22

)2

+( 𝜎₂₂− 𝜎33

)2

+( 𝜎₃₃− 𝜎11

)2]

+ 𝜎²₁₂+ 𝜎²₁₃+ 𝜎₂₃² √^𝜎 3

|𝜎|√

3 𝜏

𝑝= −¹

3

(𝜎₁₁+ 𝜎₂₂+ 𝜎₃₃)

−^𝜎

3 +^|𝜎|

3 0

3. Modeling

It has been shown frequently over the years [39,42,55,57–60], that the deformation kinetics of polymer materials can be adequately described by a linear dependence of the yield stress on the logarithm of strain rate, on temperature, and on pressure. For polymers with a thermorheologically simple response [39], an Eyring-type pressure- modified rate equation [58] can quantitatively capture the yield stress behavior:

̇̄𝛾(𝑇 , ̄𝜏, 𝑝) = ̇𝛾₀exp(

−𝛥𝑈 𝑅𝑇 )

exp (

−𝜇𝑝𝑉^∗ 𝑘𝑇

) sinh

(𝜏𝑉̄ ^∗ 𝑘𝑇

)

. (4)

Here, ̇̄𝛾 is the equivalent plastic shear rate, 𝑇 the absolute tempera- ture in 𝐾, ̄𝜏 the equivalent shear stress, 𝑝 the hydrostatic pressure,

̇𝛾₀ a rate constant, 𝛥𝑈 the activation energy, 𝑅 the universal gas constant (8.314472 J mol⁻¹K⁻¹), 𝜇 a pressure dependence parameter, 𝑉^∗the activation volume, and 𝑘 is the Boltzmann’s constant (1.38054 ⋅ 10⁻²³J K⁻¹).

The prefix ̇𝛾₀is a rate constant that reduces with diminished molecular mobility [46]. It depends on the thermodynamic state of the material and is related to the density increase due to physical ag- ing [51], which restricts molecular mobility. The first exponential term, which includes the activation energy 𝛥𝑈 , covers the material’s temperature dependence. The last term, a hyperbolic sine function that includes the activation volume 𝑉^∗, determines the stress and strain- rate dependency of the material [38,58]. The second exponential term, which includes the pressure dependence parameter 𝜇, captures the effect of hydrostatic pressure [58,60], i.e. negative for tensile tests and positive for compression tests.

Definitions for the equivalent plastic shear rate ̇̄𝛾 (based on the second invariant of the rate of strain tensor), equivalent shear stress

̄

𝜏 (based on the second invariant of the deviatoric stress tensor), and hydrostatic pressure 𝑝 are given inTable 3. Expressions for uniaxial tension, uniaxial compression, and pure shear are also given inTable 3, and show that the equivalent stress and shear rate are equal to the values measured in a pure shear test.

If written in terms of the equivalent shear stress as a function of equivalent shear rate, the equation reads:

̄

𝜏(𝑇 , ̇̄𝛾, 𝑝) =𝑘𝑇 𝑉^∗sinh⁻¹

[ ̇̄𝛾

̇𝛾₀exp(𝛥𝑈 𝑅𝑇 )

exp (𝜇𝑝𝑉^∗

𝑘𝑇 )]

. (5)

Note that true stress values are referred to in these equations.

In addition, many polymers exhibit two (or even more) molecular relaxation processes, termed as thermorheologically complex behavior.

It was illustrated previously [42,46], that PolyLactic Acid also manifests two molecular deformation processes (𝛼 + 𝛽). The effect of more than one molecular relaxation process can become visible by a change in the slope in the yield stress vs. logarithmic strain-rate plots [55]. Another way for detecting multiple relaxation processes is a possible difference between the slope of the upper-yield vs. logarithmic strain-rate and the slope of the lower-yield vs. logarithmic strain-rate in uniaxial compression tests [42,46]. In these cases, the yield behavior can be well described by the Ree–Eyring modification [37] of the Eyring-type flow rule, where it is assumed that different molecular processes act independently and in parallel:

̄

𝜏(𝑇 , ̇̄𝛾, 𝑝) =

∑𝛽 𝑥=𝛼

𝑘𝑇 𝑉_𝑥^∗sinh⁻¹

[ ̇̄𝛾

̇𝛾_0,𝑥exp (𝛥𝑈_𝑥

𝑅𝑇 )

exp

(𝜇_𝑥𝑝_𝑥𝑉_𝑥^∗ 𝑘𝑇

)]

. (6)

Here, 𝑥 relates to the indication of the parameters related to each relaxation process, 𝛼 and 𝛽, with decreasing temperature. Commonly, the molecular processes are related to a main-chain segmental motion (the primary glass- or 𝛼-transition) and a partial side-chain mobility (a secondary or 𝛽-transition) [46].

By making use of the expressions inTable 3, the pressure-dependent Ree–Eyring flow rule (Eq.(6)) can be simplified to describe the true yield stresses for both tensile as well as compression tests. This results for the uniaxial tensile tests in:

𝜎_𝑦^𝑇=

∑𝛽 𝑥=𝛼

√ 3 3 + 𝜇_𝑥

[ √3 ̇𝜀

̇𝛾_0,𝑥 exp (𝛥𝑈_𝑥

𝑅𝑇 )]

, (7)

while for the uniaxial compression tests in:

|𝜎𝑦^𝐶| =

∑𝛽 𝑥=𝛼

√ 3 3 − 𝜇_𝑥

[ √3| ̇𝜀|

̇𝛾_0,𝑥 exp (𝛥𝑈_𝑥

𝑅𝑇 )]

. (8)

Note that the only difference resides in the opposite sign for the hydrostatic pressure parameter 𝜇_𝑥. This not only accounts for a difference between the yield stress under different loading geometries (i.e. uniaxial tensile, simple shear, uniaxial compression), but also for a distinct strain-rate dependence. This last characteristic can be used to determine the influence of the hydrostatic pressure and its corresponding parameter 𝜇 [45,47,56,60].

4. Results and discussion

The PLA material used in the present study is the same as previously used [7]. From that previous study, it was concluded that the composition probably consists of a Poly(L-Lactic Acid) polymer (PLLA) with a D content of approximately 4%. The measured crystallinity of both hot-press compression molded (CM ) as well as Material Extrusion Additive Manufacturing (ME-AM) samples was in all cases below 2%.

Hence, this material’s low crystallization ability is assumed to have an insignificant effect on the mechanical properties [7].

4.1. Isotropic bulk properties

As a first step, the mechanical properties in uniaxial compression and tensile tests were determined for the CM samples. These results will help to determine some initial parameters for the Ree–Eyring flow rule, and will be considered as the reference bulk behavior of the base material. The uniaxial compression test results at various strain rates are shown inFig. 3. The stress–strain curves are the average values of three compression test results. Measurements have good repeatability as standard deviation stays below 0.95 MPa (see alsoTable A.1). As is indicated inFig. 3(a), at the highest strain rate ̇𝜀 = 1 ⋅ 10⁻¹s⁻¹the sample was affected by viscous heating at large deformations. This can be seen as a continuing reduction of the stress at higher strains, leading to a crossover with the curves at lower strain rates. Other research groups have mentioned this effect previously [46,61]. Since the Ree–

Eyring flow rule does not account for viscous heating, the lower yield data point at this strain rate was not taken into account to determine model parameters.

Fig. 3(b) demonstrates the upper and lower yield stresses, and the yield drop, as a function of logarithmic nominal strain rate. The experimental upper yield data show a distinct strain-rate dependence

(6)

Fig. 3. (a)Experimental absolute true stress/strain response in uniaxial compression tests at various strain rates for CM samples. (b) Absolute true upper (▽) and lower (△) yield stresses, and yield drop (⋆) as a function of logarithmic strain rate for CM samples. Symbols are experimental results, solid lines are model predictions. Gray symbols are data affected by viscous heating.

Fig. 4. (a)Experimental engineering stress/strain response in uniaxial tensile tests at various strain rates for CM samples. (b) Absolute true yield stresses as a function of logarithmic strain rate for CM samples. Symbols are experimental results, solid lines are model predictions. The dash–dotted line is the model prediction of the compression lower yield stress, which is related to the 𝛼-process only.

compared to the lower yield stresses. According to the method published by Van Breemen et al. [46], this is an indication of the material’s thermorheologically complex behavior. And indeed, the PLA material used in the present study can be adequately described with 2 molecular deformation processes. This was already suggested previously [7], but not experimentally demonstrated using uniaxial compression tests.

Similar behavior was found for another PLA grade [46]. The primary relaxation process, associated with the glass transition and indicated as the 𝛼-process, is assigned to cooperative local segmental motions.

The secondary or 𝛽-relaxation process is thought to be related to non-cooperative local twisting motions of the main chain [62–64].

Analogous to the PLA used in the Van Breemen study [46], the present PLA material also shows a pronounced strain softening response followed by weak strain hardening (see Fig. 3(a)). Such behavior generally leads to brittle behavior in uniaxial tensile tests. This is confirmed by the uniaxial tensile test results at various strain rates, as shown inFig. 4(a). The stress–strain curves are the average values of three tensile test results. On the other hand, the tensile test specimens show some weak stress-whitening due to the appearance of white lines perpendicular to the tensile direction distributed over the gauge length of the specimen. These lines are attributed to the crazing behavior of the PLA material [65,66], and similar to what is seen for polystyrene.

Fig. 4(b)displays the true compressive and tensile yield stresses as a function of logarithmic nominal strain rate. It can be clearly seen that the strain-rate dependence for the compressive yield stresses is

Table 4

Ree–Eyring model parameters for PLA material of CM samples.

𝑥 𝜇_𝑥 𝑉_𝑥^∗ 𝛥𝑈_𝑥^a ̇𝛾_0,𝑥^{𝑐𝑜𝑚𝑝} ̇𝛾_0,𝑥^{𝑐𝑜𝑚𝑝} ̇𝛾_0,𝑥^{𝑡𝑒𝑛𝑠} [–] [nm³] [kJ mol⁻¹] [s⁻¹] [s⁻¹] [s⁻¹]

(lower yield) (yield drop)

𝛼 0.25 4.48 480 3.81⋅ 10⁷¹ – 3.18⋅ 10⁷³

𝛽 0.25 2.46 100 – 1.25⋅ 10⁹ 1.25⋅ 10⁹

aData taken from Van Breemen et al. [46].

higher than for the tensile yield stresses. The tensile data demonstrate a slope of ±9 MPa/decade, while the compression data manifest a slope of ±12 MPa/decade (similar to but slightly lower than measured by Engels et al. [42] and Van Breemen et al. [46]). This effect is due to the influence of the hydrostatic pressure [45,58,60]. And it is used to determine the hydrostatic pressure parameter 𝜇_𝑥[45]. For illustrative reasons,Fig. 4(b)also shows the model prediction of the compressive lower yield stress, which is only determined by the primary (𝛼) molec- ular deformation process. Note that the strain-rate dependence of the 𝛼-process is different from the tensile yield stress, which is governed by both the 𝛼- and 𝛽-processes.

By making use of the methodologies as explained by Van Breemen et al. [46] and Verbeeten et al. [45], the Ree–Eyring parameters for Eqs.(7)and(8)can be determined.Table 4 gives these parameters, while the solid lines in Figs. 3 and 4show the predictive capacity

(7)

of the model. Since the transition from the (𝛼) towards the (𝛼 + 𝛽) contribution is not manifested neither in the tensile nor in the upper yield compression data, the pressure dependence parameter 𝜇_𝑥cannot be determined separately for each molecular deformation process [45]

and is taken equal for both processes. Take also into account that when the material shows a combined behavior of both the 𝛼- and the 𝛽-process, the values for the ̇𝛾_0,𝛼 and ̇𝛾_0,𝛽 parameters are not unique.

From a practical point of view, the activation volume of the 𝛼- relaxation process 𝑉_𝛼^∗determines the slope of the lower yield stress data as a function of logarithmic strain rate, while the activation volume of the 𝛽-relaxation process 𝑉_𝛽^∗determines the slope of the yield drop data.

The combination of both activation volumes determines the slope of the tensile and compression upper yield stresses as a function of logarithmic strain rate. The pressure dependence parameter 𝜇 determines the dif- ference in strain-rate dependency between the tensile and compression yield stress data. And last, the ̇𝛾_0,𝑥-parameters determine the height of the yield stress curves.

4.2. Angle-ply ME-AM samples

Compression molded samples, which are completely solid, are isotropic in nature. Its properties are mostly determined by the tem- perature and pressure profiles as a function of time, i.e. 𝑇_𝐶𝑀(𝑡)and 𝑝_𝐶𝑀(𝑡). This leads to a certain thermodynamic state of the amorphous phase and a specific amount of crystallinity (negligible for the current material [7]). Due to its isotropy, CM samples are considered as the material’s reference mechanical behavior in the present study.

On the contrary, Material Extrusion Additive Manufacturing (ME- AM) samples consist of stacked layers of partially bonded filaments with interstitial voids [9,16]. Pressure drops to atmospheric levels once the material exits the nozzle. The temperature profile is highly non-isothermal due to the initial fast cooling and successive heating cycles during the consecutive deposition of strands. And shear effects in the nozzle and due to the curvature between nozzle and deposited strand will lead to orientation and stretch of the polymer chains [11, 13,67]. Hence, mechanical properties are anisotropic and determined by a significantly different thermo-mechanical history, i.e. 𝑇_𝐴𝑀(𝑡)and

̇𝛾_𝐴𝑀(𝑡).

Furthermore, the ME-AM voided structure has an influence on the measured stress–strain curves, as voids lower the stress values. To be able to compare different 3D printed sets, which generally have a variation in porosity, between each other and with the CM sample set, the voided structure is compensated using Eq.(3)to obtain ‘‘void corrected’’ or ‘‘solid’’ yield stress results. Thus, the effect of the infill orientation stacking sequence can be determined in a macroscopic sense.

The stress–strain results for the symmetric angle-ply ME-AM sample sets are given inFig. 5Results for the quasi-isotropic set is also shown in Fig. 5. Average porosities and apparent densities for each set are given inTable 5. The shown stress–strain curves are average curves of three tensile test results. In the subfigures on the right, the directly measured yield stresses are indicated with open symbols, while the void corrected results are displayed by closed symbols.

As was also shown previously [7], it immediately draws the attention that with the current printing parameters all ME-AM sample sets show significantly more ductile behavior than the CM sample set (2.6%

strain at break for the CM sample, and ranging from 3.1% to 6.6% for the ME-AM samples at a strain rate of ̇𝜀 = 10⁻³s⁻¹; see alsoTable A.2).

This is a clear effect of the ME-AM processing step. Additionally, ductility seems to be higher for the intermediate infill orientations, i.e. sample set PLAo45 and PLAo60. Stress-whitening in the form of white lines perpendicular to the tensile force direction are observed in all test samples. The amount of stress-whitening, however, is higher than for the CM test samples. Furthermore, samples which show higher strain-at-break also exhibit more stress-whitening. This indicates that multiple craze formation seems to be the failure mechanism for this

Table 5

Average apparent density, porosity, and dimensional length change of ME-AM samples.

Standard deviations are indicated between brackets.

Sample 𝜌̄_app Porosity Length

nomenclature [g cm⁻³] [%] [%]

PLAo00 1.12 (0.009) 10.1 (0.7) −12.8 (0.9)

PLAo15 1.17 (0.009) 6.6 (0.8) −8.8 (0.8)

PLAo30 1.16 (0.006) 7.5 (0.5) −7.2 (0.5)

PLAo45 1.16 (0.006) 7.0 (0.5) −5.8 (0.7)

PLAo60 1.16 (0.007) 7.2 (0.5) −4.2 (0.2)

PLAo75 1.14 (0.004) 9.1 (0.3) −3.2 (0.1)

PLAo90 1.15 (0.006) 7.8 (0.4) −3.3 (0.1)

PLAoISO 1.14 (0.005) 8.5 (0.4) −5.5 (0.6)

Table 6

Model parameters used to describe the yield behavior of Compression Molded (CM ) and ME-AM samples. Pressure dependence parameters are taken as 𝜇_𝛼= 𝜇_𝛽 = 0.25.

Activation energies are taken as 𝛥𝑈_𝑎=480 kJ mol⁻¹and 𝛥𝑈_𝑏=100 kJ mol⁻¹[46].

Sample 𝑉_𝛼^∗ ̇𝜀_0,𝛼 𝑉_𝛽^∗ ̇𝜀_0,𝛽

nomenclature [nm³] [s⁻¹] [nm³] [s⁻¹]

CM 4.48 3.92⋅ 10⁶⁸ 2.46 6.23⋅ 10¹¹

PLAo00 4.48 2.49⋅ 10⁶² 2.46 5.95⋅ 10¹⁴

PLAo15 4.48 4.90⋅ 10⁶² 2.46 1.28⋅ 10¹⁵

PLAo30 4.48 5.81⋅ 10⁶² 2.46 1.62⋅ 10¹⁵

PLAo45 4.48 5.39⋅ 10⁶² 2.46 2.14⋅ 10¹⁵

PLAo60 4.48 9.80⋅ 10⁶² 2.46 2.10⋅ 10¹⁵

PLAo75 4.48 4.63⋅ 10⁶³ 2.46 2.43⋅ 10¹⁵

PLAo90 4.48 4.55⋅ 10⁶³ 2.46 2.98⋅ 10¹⁵

PLAoISO 4.48 9.92⋅ 10⁶² 2.46 3.45⋅ 10¹⁵

PLA material [65,66], also holding up for ME-AM processed samples.

The SEM fractography section will provide more details on the fracture mechanism.

The calculated apparent densities of the current sample sets are lower than the ones calculated in the previous study [7]. As a consequence, sample porosities are higher. While in the previous study all average porosities were below 6% [7], in the present research values range between 6.6% and 10.1%. Although the same printer parameters are used, a different printer is employed, which has its effect on the measured results. Nevertheless, the trends shown by the results are very similar.

The strain-rate dependence of the yield stresses for these ME-AM sample sets can be adequately captured using the pressure-dependent Ree–Eyring modification of the Eyring-type flow rule as given by Eq.(7). The model predictions are shown in the subfigures on the right ofFig. 5as black solid lines. Model parameters are given inTable 6. Ex- cellent model descriptions are obtained, especially taken into account that the same activation volumes for the 𝛼- and 𝛽−relaxation processes as determined from the CM sample results are used for all ME-AM sample sets (seeTable 6). The coefficient of determination 𝑅² is for all sets above 0.98, demonstrating the outstanding correlation between experiments and model. Except for set PLAo90, where the coefficient of determination drops slightly to 𝑅²≈ 0.95, still manifesting a satisfying fit.

Note that the values for the rate constants ̇𝜀_0,𝛼 and ̇𝜀_0,𝛽 of the CM sample set inTable 6are different from the values given inTable 4. The rate constant for the 𝛼-deformation process is lower, while it is higher for the 𝛽-process. However, the model performance is completely equal for both parameter sets. This confirms the argument that values for the rate constants are not unique, if no distinction can be made between both deformation processes.

Also remarkable is the distinct change in slope of the strain-rate dependence of the yield stress inFig. 5. The ME-AM sample results evidently show a transition from a low strain-rate region with only the contribution from a single deformation process (𝛼) towards a high strain-rate region with contributions of two deformation processes (𝛼 + 𝛽). This clear transition is not detected for the compression molded tensile test specimen as shown inFig. 4(b).

(8)

Fig. 5. Engineering stress–strain response in uniaxial tensile tests at various strain rates for ME-AM samples. (a,c,e,g,i,k,m,o) Stress as a function of strain. Symbols are experimental yield stresses. (b,d,f,h,j,l,n,p) Yield stress and void corrected yield stress as a function of logarithmic strain rate. Symbols are experimental results, solid lines are model predictions.

(a,b)[0^◦], (c,d) [15^◦/−15^◦], (e,f) [30^◦/−30^◦], (g,h) [45^◦/−45^◦], (i,j) [60^◦/−60^◦], (k,l) [75^◦/−75^◦], (m,n) [90^◦], (o,p) [0^◦/90^◦/45^◦/−45^◦].

Although a transition from an 𝛼-region towards an 𝛼 + 𝛽-region was also suggested in the previous study [7], experimental evidence was less clear. The reason may be that a relatively low number of strain rates was applied. Here, more strain rates are measured over the same strain- rate range, and the transition has become more apparent. A plausible explanation for this transition is suggested in the next subsection.

4.3. Orientation induced anisotropy

Several research groups have mentioned, reasoned, and measured that the ME-AM processing technique induces local orientation and stretch of the molecular chains [7,11,13,16,18,67,68]. It was also previously shown that variations in molecular chain orientation and stretch

(9)

Fig. 5. (continued).

produce changes in mechanical properties, both for ME-AM components [7,16], as well as oriented polymers due to hot drawing [69,70], hydrostatic extrusion [69,71], or even injection molding [53,54].

Using a thermal shrinkage procedure [7,16], a qualitative macroscopic measurement for molecular orientation and stretch can be determined. Dimensional variations are due to, on the one hand, cold crystallization and, on the other hand, relaxation of the residual chain alignment. As explained in Section2.6, the variations in dimensional changes are an indication for the molecular chain relaxation.

Table 5shows the average dimensional length changes of the var- ious ME-AM sample sets. Logically, the PLAo00 set has the highest

length contraction (12.8%), as the deposited polymer strands are oriented completely in the sample length direction. Augmenting the orientation angle from [0^◦] to [90^◦] reduces this length change, towards the minimum of 3.2% contraction for both PLAo75 and PLAo90 sets.

The quasi-isotropic PLAoISO sample set has a contraction close to but slightly lower than the PLAo45 sample set. Therefore, the most pronounced effect on the mechanical properties, compared to the CM samples, should be visible for the PLAo00 sample set, as the tensile forces are imposed in the direction of the deposited strands.

Fig. 6compares the strain-rate dependence of the (void corrected) yield stress between the CM and PLAo00 sample sets. It can be easily seen that the contribution of the 𝛼-deformation process has increased

(10)

Fig. 6. Yield stress and void corrected yield stress as a function of logarithmic strain rate for the CM and PLAo00 sample sets. Symbols are experimental data, lines are model predictions.

significantly for the PLAo00 samples compared to the CM set. As a consequence, only this 𝛼-relaxation process becomes noticeable at low strain rates, while both molecular relaxation processes (𝛼 + 𝛽) become visible at higher strain rates. An (𝛼 → 𝛼 + 𝛽)-transition is provoked for the PLAo00 sample set, which is not detected for the CM sample set. When the rate constants ̇𝜀_0,𝛼 and ̇𝜀_0,𝛽fromTable 6are compared, the ̇𝜀_0,𝛼 parameter diminishes for PLAo00, while ̇𝜀_0,𝛽 increases. This implicates that the primary 𝛼-process, assigned to cooperative local segmental motions of the molecular backbone, is hindered, whereas the secondary 𝛽-process, related to non-cooperative local twisting motions of the main chain, becomes somewhat easier. Hence, it is suggested that the oriented and extended molecular chain as a result of the ME- AM processing step (note the 12.8% length contraction for this sample set in Table 5), hampers the 𝛼-deformation process of the molecular backbone. As a consequence of this reduced molecular mobility, yield stresses related to that primary relaxation process increases. Subse- quently, deformation related to the 𝛽-process is slightly favored and yield stresses, attributed to this secondary process, diminish. A similar response was indicated by Senden et al. [72] for oriented polycarbonate (PC).

As the orientation angle is increased, the chain orientation and ex- tension rotates away from the axis of the tensile forces during mechanical characterization (compare the length contraction data inTable 5).

To show how it affects the yield stresses, the ME-AM sample sets at different infill orientations are compared inFig. 7(a). All sets still show the transition from the 𝛼-relaxation towards the (𝛼+𝛽)-relaxation. How- ever, curves move down and to the right at increasing infill orientation angle. Although still hindered, molecular chain deformation seems to become easier at increasing angle. As a consequence, the rate constants for the 𝛼- and 𝛽-processes slowly increase, as the values fromTable 6 indicate.

Senden et al. [72], in a study on oriented polycarbonate and in a subsequent study on oriented polyethylene (PE) [54], explored and modeled anisotropy in oriented polymers by focusing on loading angle and strain rate. Their model featured two sources of anisotropy: a frozen-in elastic network stress due to initial material orientation in combination with an anisotropic viscoplastic flow rule based on a Hill yield function. A frozen-in internal stress, as a consequence of a pre-stretched elastic network, provokes increased strain hardening in tensile tests. Such effect may also explain the more ductile behavior of ME-AM processed samples in the tensile test experiments of the present study. Furthermore, the behavior shown inFig. 7(a)displays similar tendencies and features as results from Senden et al. [54].

different infill orientation angles can be adequately described with a Ree–Eyring flow rule using 2 fixed activation volumes (as determined on CM processed samples) and changing rate constants. To analyze these rate constants, both ̇𝜀_0,𝛼- and ̇𝜀_0,𝛽-parameters are shown as a function of the infill orientation angle inFig. 8.Fig. 8(a)demonstrates that the rate constant related to the 𝛼-process stays almost constant at both low and high infill orientation angles. In between, at around approximately 65^◦, the plot exhibits a kind of ‘‘jump’’ from a lower to a higher value. On the other hand, Fig. 8(b) seems to reveal a steady increase of the ̇𝜀_0,𝛽-parameter below an infill orientation angle of 60^◦, followed by a slower change at higher orientation angles. This insinuates that at lower orientation angles, it is easier for the material to deform due to twisting motions instead of local segmental motions.

At higher orientation angles, when the oriented and stretched polymer is rotated far enough away from the tensile force axis, the cooperative local segmental motions start to play a more important role in the material’s deformation dynamics.

If the tendencies as shown inFig. 8are taken into account for the rate constants in the Ree–Eyring modification of the Eyring flow rule (i.e. the dashed lines), the yield stresses at different strain rates and as a function of orientation angle can be predicted.Fig. 9 shows the predictions at a lower strain rate and at a higher strain rate. At the strain rate of ̇𝜀 = 10⁻⁴s⁻¹, mostly the 𝛼-relaxation process is active.

At the higher strain rate of ̇𝜀 = 10⁻²s⁻¹, both deformation processes act. For the lower strain rate, yield stresses are slightly underpredicted.

However, the tendency of the yield stress as a function of orientation angle is clearly well captured. For the higher strain rate, yield stresses are excellently modeled. The other strain rates, although not shown here, slightly over- or underpredict the yield stress values, similar to Fig. 9(a). Nevertheless, the trend is definitely well captured. Note that the range of the 𝑦-axes in Fig. 9 is smaller than for Fig. 5, which enlarges the differences between the measured data and the model.

As can be concluded from the previous results, this PLA material manifests a deformation-dependent Eyring rate constant. Its activation volumes for the two relaxation processes can be adequately determined on compression molded samples for tensile and compression characterization tests and do not seem to change due to different processing techniques. The rate constants, however, are determined by the thermo- mechanical history (i.e. time–temperature and time–strain profiles) the material undergoes during the ME-AM process. This was also concluded in a previous study [7]. Nevertheless, this does not hold for all ME-AM processed materials. For example, it was shown that ME-AM processed ABS rather shows a deformation-dependent activation volume [18].

4.4. SEM fractography

Following uniaxial tensile testing, the fracture surface of the as- received filament, hot-press compression molded specimens, and material extrusion additive manufactured test samples were observed using scanning electron microscopy. Observations provide additional insight on the fracture mechanism for this PLA material.

The as-received PLA filament was characterized under tension at a strain rate of ̇𝜀 = 10⁻²s⁻¹until failure. A cross-section of the filament’s fractured surface was analyzed by SEM micrography and is shown in

(11)

Fig. 7. (a) Void correctedyield stress as a function of logarithmic strain rate for various ME-AM sample sets. (b) Void corrected yield stress as a function of logarithmic strain rate for the quasi-isotropic ME-AM sample set compared to PLAo00 and PLAo90. Symbols are experimental results, solid lines are model predictions.

Fig. 8. Rate constant ̇𝜀0as a function of infill orientation angle. (a) Related to 𝛼-relaxation process. (b) Related to 𝛽-relaxation process. Symbols are values fromTable 6, dashed lines are a guide to the eye.

Fig. 9. Volume correctedtrue yield stress as a function of infill orientation angle. (a) For strain rate ̇𝜀 = 10⁻⁴s⁻¹. (b) For strain rate ̇𝜀 = 10⁻²s⁻¹. Symbols are experimental values, solid lines are model predictions.

Fig. 10. The filament fractures in a macroscopically brittle manner, as can be deduced from Fig. 10(a). The surface is rather smooth and only few plastically deformed regions can be seen, indicated by white zones/lines. These are, however, very localized. Especially the filament border demonstrates these plastic deformation zones, as shown inFig. 10(b).

At magnifications of 200x and higher very thin lines of material become visible, which are highly extended micro-fibrils. These can be detected, among others, near the tip of the green arrow inFig. 10(b). At a higher magnification, seeFig. 10(c), such a micro-fibril has recoiled onto the surface. As this image shows, one end of the extended material is attached to the surface and shows high ductile tearing, while the

(12)

Fig. 10. SEM micrographs of the as-received PLA filament cross-section. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 11. SEM micrographs of the CM PLA sample fracture cross-section. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. SEM micrographs of the ME-AM PLAo00 sample fracture cross-section, [0^◦]. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

other end is lying loose on the surface. This seems to suggest that micro- fibrils start during mechanical characterization, are extended during the test, and break at the moment of sample fracture. Hence, as was suggested by Jiang et al. [73], it seems likely that these micro-fibrils are formed during crazing of the material.

In Figs. 11 to 15, SEM micrographs of the fracture surfaces of arbitrary samples for the different sample sets are displayed. All shown samples were mechanically characterized at a strain rate of ̇𝜀 = 10⁻²s⁻¹, except for the CM sample, which was tested at ̇𝜀 = 10⁻³s⁻¹. Although not shown here, no significant variations on the fracture surfaces were observed for samples that were tested at other strain rates.

A cross-section of a mechanically characterized CM PLA sample analyzed by SEM micrography is shown in Fig. 11. The left half of the fractured sample demonstrates a more brittle fracture surface with rather few white zones (related to plastic deformation), as shown in Fig. 11(a). On the other hand, the right half shows more intense white stripes indicating more plastically deformed regions. The fracture surface is still typical for brittle fracture, though. The direction of the white lines on the right half seem to indicate that the fracture initiated on the left side of the sample. Note thatFig. 4(a)manifests macroscopically brittle behavior for this sample set.

At a higher magnification, i.e. Fig. 11(b), some micro-fibrils become visible (near the tip of the green arrows), similar to the PLA filament images. Although there are some zones with a higher amount of micro-fibrils, the density of micro-fibrils is rather low on this CM sample fracture surface, compared to the ME-AM surfaces. As mentioned before, these micro-fibrils are thought to be related to crazing behavior [66,73].

InFig. 11(c), a high magnification of 10,000x is shown. Apart from the white zone with local plastic deformation, small micro-cracks are visible near the tip of the yellow arrow. AlthoughFig. 11(c)suggests that they seem to be present in a large portion of the fracture surface, these micro-cracks only become visible at this high magnification of 10,000x. It is suggested that these micro-cracks are the reason for the stress-whitening seen in the gauge section of the tensile test specimen at a macroscopic scale, and, hence, thought to be related to crazing.

The images related to the surfaces of ME-AM processed samples, Figs. 12 to 15, all show brittle surface fracture. But, the amount of plastic deformation (white zones) is higher compared to the CM sample.Figs. 12(b)and12(c)more clearly show that the white zones are related to local ductile tearing of the material.

Especially inFig. 12, which displays representative SEM images of the fracture surface of an ME-AM PLAo00 sample, triangular-shaped inter-strand and inter-layer voids can be easily detected. A few voids,